Extreme Y Chromosome Polymorphism

Home , Poecilia parae, Poeciliidae

bioRxiv preprint doi: https://doi.org/10.1101/2020.08.19.258434; this version posted August 20, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-ND 4.0 International license.

1 Extreme Y chromosome polymorphism corresponds to five male reproductive 2 morphs 3 4 5 Benjamin A Sandkama*, Pedro Almeidab, Iulia Daroltia, Benjamin Furmana, Wouter van 6 der Bijla, Jake Morrisb, Godfrey Bournec, Felix Bredend, Judith E. Manka,b

7 a. Department of Zoology, University of British Columbia, Vancouver, BC V6T 1Z4, 8 Canada

9 b. Department of Genetics, Evolution and Environment, University College London, 10 London WC1E 6BT, United Kingdom

11 c. Department of Biology, University of Missouri-St. Louis, St. Louis, MO 63105, USA

12 d. Department of Biological Sciences, Simon Fraser University, Burnaby, BC V5A 1S6, 13 Canada

14 15 * Corresponding Author: Benjamin A Sandkam 16 Email: [email protected] 17 18 Keywords 19 Genome Evolution, Alternative Mating Tactics, Poecilia parae, Supergene 20 21 Author Contributions 22 B.A.S. and J.E.M. designed research; B.A.S., J.E.M, F.B., G.R.B. conducted field work; 23 B.A.S., P.A., I.A., B.L.S.F., W.v.d.B., and J.M. conducted bioinformatic analyses; B.A.S., 24 P.A., I.A., B.L.S.F., W.v.d.B., J.M., G.R.B., F.B. and J.E.M. wrote the paper. 25 26 27

28 Abstract 29 Sex chromosomes form once recombination is halted between the X and Y 30 chromosomes. This loss of recombination quickly depletes Y chromosomes of 31 functional content and genetic variation, which is thought to severely limit their potential 32 to generate adaptive diversity. We examined Y diversity in Poecilia parae, where males 33 occur as one of five discrete morphs, all of which shoal together in natural populations 34 where morph frequency has been stable for over 50 years. Each morph utilizes different 35 complex reproductive strategies, and differ dramatically from each other in color, body 36 size, and mating behavior. Remarkably, morph phenotype is passed perfectly from 37 father to son, indicating there are five Y haplotypes segregating in the species, each of 38 which encodes the complex male morph characteristics. Using linked-read sequencing 39 on multiple P. parae females and males of all five morphs from natural populations, we 40 found that the genetic architecture of the male morphs evolved on the Y chromosome 41 long after recombination suppression had occurred with the X. Comparing Y 42 chromosomes between each of the morphs revealed that although the Ys of the three 43 minor morphs that differ predominantly in color are highly similar, there are substantial 44 amounts of unique genetic material and divergence between the Ys of the three major 45 morphs that differ in reproductive strategy, body size and mating behavior. Taken 46 together, our results reveal the extraordinary ability of evolution to overcome the 47 constraints of recombination loss to generate extreme diversity resulting in five discrete 48 Y chromosomes that control complex reproductive strategies.

49 Significance Statement 50 The loss of recombination on the Y chromosome is thought to limit the adaptive 51 potential of this unique genomic region. Despite this, we describe an extraordinary case 52 of Y chromosome adaptation in Poecilia parae. This species contains five co-occurring 53 male morphs, all of which are Y-linked, and which differ in reproductive strategy, body 54 size, coloration, and mating behavior. The five Y-linked male morphs of P. parae 55 evolved after recombination was halted on the Y, resulting in five unique Y 56 chromosomes within one species. Our results reveal the surprising magnitude to which 57 non-recombining regions can generate adaptive diversity and have important

58 implications for the evolution of sex chromosomes and the genetic control of sex-linked 59 diversity.

60 Introduction

61 Sex chromosomes form when recombination is halted between the X and Y 62 chromosomes. The loss of recombination results in a host of evolutionary processes 63 that quickly deplete Y chromosomes of functional content and genetic variation, 64 severely limiting the scope for adaptive evolution1. Y chromosomes can counter this 65 loss to some degree through a variety of mechanisms2-5, however the adaptive potential 66 of Y chromosomes is generally thought to be much lower than the remainder of the 67 genome. Typically, this results in relatively low levels of Y chromosome diversity within 68 species. The adaptive potential of non-recombining regions has far broader implications 69 beyond just Y chromosomes given the increasing realization that supergenes, linked 70 regions containing alleles at multiple loci underlying complex phenotypes, are key to 71 many adaptive traits6-12. Many supergenes are lethal when homozygous and therefore 72 also non-recombining8,10. Therefore, the processes that constrain Y chromosome 73 evolution also affect much broader areas of the genome.

74 Poecilia parae is a small freshwater fish found in coastal streams of South America. 75 Remarkably, males of this species are one of five distinct morphs that utilize different 76 reproductive tactics, and differ in reproductive strategy, body size, color, and mating 77 behaviour13-19 (summarized in Table S1). There are three major morphs: parae, 78 immaculata and melanzona. The parae morph has the largest body, vertical black 79 stripes, an orange tail-stripe and is highly aggressive, chasing away rival males and 80 aggressively copulating with females by force. Immaculata, resembles a juvenile 81 female. Although immaculata has the smallest body size of all morphs, it has the largest 82 relative testes and produces the most sperm, employing a sneaker copulation strategy. 83 Melanzona is sub-divided into three minor morphs, which are similar in body size and all 84 have a colored horizontal stripe (either red, yellow or blue), which they present to 85 females during courtship displays.

86 All five morphs co-occur in the same populations, and the relative frequency of morphs 87 is highly stable over repeated surveys spanning 50 years (~150 generations)13,14,20. This 88 suggests that balancing selection, likely resulting from a combination of sexual and 89 natural selection, is acting to maintain these five adapted morphs. Most importantly,

90 multigeneration pedigrees show that morph phenotype is always passed perfectly from 91 father to son13, indicating morphs are Y-linked. Given their discrete nature and complex 92 phenotypes, it is clear that the five P. parae morphs are controlled by five different Y 93 chromosomes. This system therefore offers the potential for a unique insight into the 94 adaptive potential of Y chromosomes, and the role of these regions of the genome in 95 male phenotypes.

96 We have recently shown that poecilid species closely related to P. parae share the 97 same sex chromosome system as Poecilia reticulata21 (guppies), however the extent of 98 Y chromosome degeneration differs markedly across the clade. Although the Y 99 chromosome in P. reticulata and Poecilia wingei contains only a small area of limited 100 degeneration21-24, the entirety of the Y chromosome of Poecilia picta is highly 101 degenerate21. P. parae is a sister species of P. picta, however P. picta males are 102 markedly different from P. parae and do not resemble any of the five P. parae 103 morphs18,20,25-27, suggesting extreme diversity was generated on the P. parae Y 104 chromosome after recombination was halted with the X chromosome. Work on model 105 systems has indeed shown Y chromosomes can accumulate new genetic material2-5, 106 yet these differences occur over long periods of time and are only evident when 107 comparing Ys across species. Non-model systems, such as P. parae, provide a unique 108 opportunity to explore the limits and role of non-recombining regions in generating 109 diversity.

110 Because P. parae is very difficult to breed in the lab we collected tissue from natural 111 populations in South America where all five male morphs co-occur and used linked-read 112 sequencing on multiple females and males of all five morphs. We first confirmed that P. 113 parae shares the same sex chromosome system as its close relatives21,22. We went on 114 to find patterns of X-Y divergence are the same for all five Y chromosomes and 115 matches the X-Y divergence we observed in P. picta, suggesting that the morphs 116 emerged after Y chromosome recombination was stopped in a common ancestor of P. 117 parae and P. picta. Comparing the five Y chromosomes to each other, we find that while 118 the Ys of the three minor morphs (red, yellow and blue melanzona) that differ only in 119 color are highly similar, the Ys of the three major morphs (parae, immaculata, and 120 melanzona) that differ in reproductive strategy, body size and mating behavior are 5

121 significantly diverged from one another and carry substantial amounts of unique genetic 122 material. Taken together, our results reveal the surprising ability of the Y chromosome 123 to not only overcome the constraints of recombination loss, but to generate extreme 124 diversity, resulting in five discrete Y chromosomes that control complex reproductive 125 strategies.

126

127 Results

128 We collected 40 individual P. parae from natural populations in Guyana in December 129 2016, including eight red melanzona, four blue melanzona, five yellow melanzona, five 130 immaculata, seven parae morph males, and 11 females. 29 samples with sufficiently 131 high molecular weight were individually sequenced with 10X Genomics linked-reads. 132 We generated a de novo genome assembly for each of these samples. The remaining 133 11 lower molecular weight samples were individually sequenced with Illumina 134 sequencing paired end reads (see Tables S2 and S3 for sequencing and assembly 135 details).

136

137 The P. parae Y chromosome is highly diverged from the X and shared with P. 138 picta

139 Degeneration of the Y chromosome results in reduced male coverage when mapped to 140 a female reference genome. The ratio of male to female mapped reads can be used to 141 identify regions where the Y and X chromosomes differ substantially from each 142 other21,28-30. To do this, we used our best female de novo P. parae genome, based on 143 N50 and other assembly statistics (see Table S3). We then determined chromosomal 144 position of the scaffolds using the reference-assisted chromosome assembly (RACA) 145 pipeline, which combines phylogenetic and sequencing data to place scaffolds along 146 chromosomes31. Next we mapped reads from all 40 samples to this female assembly 147 and calculated male:female coverage, first for each of the five morphs independently, 148 and then all morphs together.

149 As we previously found in P. picta21, chromosome 8 (syntenic to P. reticulata 150 chromosome 12) showed a clear signal of reduced read coverage in males (Figure 1a- 151 b), indicating an XY sex determination system. Y divergence is evident across nearly 152 the entire chromosome and is largely identical to the pattern we previously observed in 153 P. picta (Figure 1c and Fig. S1). This suggests that these species inherited a highly 154 degenerate Y chromosome from their common ancestor, well before the origin of the P. 155 parae male morphs.

156 Short sequences representing all the possible substrings of length k that are contained 157 in a genome are referred to as k-mers, and k-mer comparisons between male and 158 female genomes has been used to identify Y chromosome sequence (Y-mers) in a wide 159 range of organisms32-34, including guppies35 and P. picta21. We first compared all males, 160 representing all five morphs, to all females. We found a total of 27,950,090 Y-mers (of 161 31bp) that were present in at least two males but absent from all females. However, 162 only 59 of these Y-mers were present in all 23 males (Figure 2). We found only 251,472 163 k-mers present in at least two females but absent from all males, and 0 of these were 164 found in all 6 females, demonstrating our Y-mer approach had a very low false positive 165 rate.

166 We next used Y-mer analysis to further test whether recombination was halted on the Y 167 in the common ancestor of P. parae and P. picta. Of the 646,745 Y-mers that we 168 previously identified in at least one P. picta male and no females21, 790 P. picta Y-mers 169 matched Y-mers we identified in P. parae, consistent with a shared history of 170 suppressed recombination. Additionally, these shared Y-mers were present in males of 171 all morphs (Fig. S2) and discussed in more detail below. These shared Y-mers, 172 combined with the striking similarity in male:female read mapping (Fig. S1) provide 173 compelling evidence that the vast majority of Y chromosome recombination suppression 174 occurred in a common ancestor of P. picta and P. parae.

175

176 The P. parae Y chromosomes are highly diverged from each other

177 We next compared Y-mers across individuals, generating a phylogeny on the 178 presence/absence of Y-mers in all the P. parae individuals with P. picta as an outgroup

179 (Figure 2). Clear clades were recovered for each of the major morphs (immaculata, 180 parae and melanzona) while the three minor morphs of melanzona (red, yellow, blue) 181 were very similar to one another.

182 The phylogenic relationships of individuals (Figure 2) closely match the relative Y-mer 183 comparisons across morphs (Figure 3). We found 64,515 Y-mers in every immaculata 184 male that were not in any parae or melanzona males (i.e. immaculata-mers), 87,629 185 melanzona-mers, and 1,435 parae-mers, suggesting that the melanzona and 186 immaculata Y chromosomes may contain more unique, non-repetitive sequence 187 compared to the parae Y. Moreover, we found 10,673 Y-mers in all melanzona and 188 parae males that did not occur in any immaculata males (Figure 2), suggesting that the 189 parae Y shares greater sequence similarity to the melanzona Y.

190 We calculated our false positive rate by randomly permuting our male samples into 191 groups regardless of morph and determining Y-mers present in all males of each group 192 that were absent from all other males. We found no unique Y-mers in groups of five or 193 more random males, and just 31 unique Y-mers in groups of four random males, 194 demonstrating the false positive rate of our morph-mer approach is exceedingly low 195 (Fig. S3).

196

197 Mapping morph-mers confirms high diversity of Y chromosomes

198 The large number of morph-mers we identified could either indicate that the discrete 199 morphs are the result of low divergence across large Y chromosomes, or smaller 200 complexes of highly diverged Y sequence. To resolve this, we mapped the respective 201 set of morph-mers to the 21 de novo male genomes. If divergence across the morph- 202 specific Y chromosomes is low compared to each other, mapped morph-mers would be 203 dispersed across many scaffolds. Instead we found morph-mers were not evenly 204 dispersed. For example, a single melanzona scaffold (~110kb) containing 27% of all 205 melanzona-mers (23,773), and most morph-mers overlapped one another (Fig. S4 and 206 S5). This confirms our morph-mer approach identified complexes of highly diverged Y 207 sequence.

208 To compare the relative size of these diverged complexes across morphs, we identified 209 all scaffolds that contained >5 morph-mers in each individual. The average amount of 210 sequence contained within these morph-mer scaffolds was 1.3 Mb for melanzona, 3.2 211 Mb for immaculata, and just 0.1 Mb for parae individuals (Table S4). This complements 212 the relative number of Y-mers we found for each morph and together suggests the 213 amount of unique Y chromosome sequence differs across morphs, with the parae 214 morph Y containing the smallest amount of unique genetic material.

215

216 Read mapping confirms high divergence of Y chromosomes

217 To determine how divergent the five Y chromosomes are from one another, we mapped 218 reads from all 39 samples to full de novo genome assemblies of each male morph (200 219 total alignments). Most scaffolds contain autosomal sequence, and coverage is not 220 expected to differ by sex or morph. Meanwhile, scaffolds containing morph specific 221 sequence will have higher coverage by males of the same morph (e.g. immaculata 222 reads mapped to an immaculata assembly) than coverage by males of a different morph 223 (e.g. melanzona reads mapped to an immaculata assembly). Low female coverage of 224 such scaffolds confirms these regions are on the Y and are substantially diverged from 225 the X.

226

227 As expected, when comparing coverage between males of the same morph as the 228 reference assembly and the other morphs we found average coverage was 1:1 when 229 considering all scaffolds, yet scaffolds enriched for morph-mers (containing >5) had 230 much higher coverage by males of the same morph as the reference (Figure 4). 231 Surprisingly, we found no coverage by immaculata or parae reads for nearly half 232 (40/100) of the scaffolds enriched for melanzona-mers. Similarly, 14 of the 93 scaffolds 233 enriched for immaculata-mers had no coverage when we mapped melanzona and 234 parae reads. Meanwhile, in agreement with our morph-mer analysis, all 12 of the 235 scaffolds enriched for parae-mers had nearly equal coverage by melanzona and 236 immaculata reads, once again suggesting that the parae Y contains very little unique Y 237 sequence.

238 We also found the ratio of male:female coverage was much higher for morph-mer 239 scaffolds, many of which had no female coverage, again confirming these complexes of 240 morph specific sequence are located in non-recombining regions of the Y chromosome 241 (Figure 4).

242

243 Gene annotation of morph-mer scaffolds

244 We identified genes on scaffolds with >5 morph-mers. In total, we found 7 genes on the 245 scaffolds containing the 59 Y-mers present in all morphs (totaling 30,558,901 bp), 291 246 genes on the immaculata scaffolds of sample P09 (totaling 9,748,162bp), 15 genes on 247 the melanzona scaffolds of sample P01 (totaling 295,057 bp), and no genes on the 248 parae morph scaffolds of P04 (totaling 127,542 bp) (Tables S5, S6 and S7).

249 Only one gene was predicted on scaffolds that were completely unique to melanzona 250 (trim35), and only two genes were predicted on scaffolds completely unique to 251 immaculata (trim39 and nlrc3). Members of the Trim gene family act throughout the 252 body and are well known to rapidly evolve novel functions36,37. Meanwhile, nlrc3 has 253 been shown to selectively block cellular proliferation and protein synthesis by inhibiting 254 the mTOR signalling pathway38, which could play a role in keeping immaculata the 255 smallest morph. Several copies of the transcription factor Tbx3 were present on male 256 unique scaffolds in both melanzona and immaculata morphs. Tbx genes play key roles 257 in development and act as developmental switches39-41, raising the possibility that it 258 could play a role in orchestrating the multi-tissue traits that differ across morphs.

259 We also found several copies of texim genes on male scaffolds of melanzona and 260 immaculata that most closely match texim2 and texim3. While the function of texim2 261 and texim3 are largely unknown they have been shown to be highly expressed in the 262 brain and testis of closely related species42. In Xiphophorus maculatus, a close relative 263 to P. parae (~45 mya43), the transposable element helitron has moved texim1 to the sex 264 determining region of the Y chromosome and duplicated it resulting in three copies of 265 texim that are expressed specifically in late stage spermatogenesis42. The copies of 266 texim2 and texim3 that we identified on the Y chromosome of P. parae are not the same

267 as those in X. maculatus as the Ys arose independently and the X. maculatus Y is not 268 chromosome 8.

269 We also found a large number of transposable elements on scaffolds enriched for the Y- 270 mers that were present in all individuals, and the scaffolds enriched for morph-mers, 271 including 90 copies of the helitron transposable element on scaffolds enriched for 272 melanzona-mers, and 38 copies on scaffolds enriched for immaculata-mers (Tables S9 273 and S10). It is possible that a process similar to X. maculatus has occurred in P. parae 274 where either TEs have moved texim2 and texim3 to Y specific sequence or Y specific 275 sequence has evolved around these genes. Future analyses are needed to determine 276 the roles of these and other genes in generating the morph-specific phenotypes.

277

278 Discussion

279 Recombination is widely regarded as one of the most important processes generating 280 phenotypic diversity as it produces novel allelic combinations on which selection can 281 act44. The loss of recombination is classically assumed to prevent the generation of 282 novel large-scale phenotypes, as non-recombining regions are expected to rapidly lose 283 diversity through sweeps and background selection1,45-53 and have only limited potential 284 for adaptive evolution. The power of these processes to deplete non-recombining 285 regions of diversity is clearly evident in the Y chromosomes of many species1.

286 Our results, based on both similarity of Y degeneration (Fig. S1) and shared Y-mers, 287 are consistent with recombination suppression between the X and Y chromosome in the 288 common ancestor of P. parae and P. picta (14.8-18.5 mya43) and a highly degenerate Y 289 present at the origin of P. parae. Importantly, none of the characteristics that 290 differentiate the immaculata, parae or three melanzona morphs of P. parae are found in 291 any close relatives, which means the genetic basis of the extreme diversity in morphs 292 evolved on a non-recombining, highly degenerate Y chromosome. Given that these 293 morphs differ in a suite of complex traits, including body size, testis size, color pattern, 294 and mating strategy, it is highly likely that they are underpinned by polygenic genetic 295 architectures, which evolved on the P. parae Y chromosome. Consistent with complex

296 differences between morphs, we identified substantial morph-specific genetic material 297 (Figures 2-4) that was also absent in females and therefore Y-linked.

298 Although male-specific regions of the genome may experience elevated mutation 299 rates54, it is far more likely that the high P. parae Y diversity was generated through 300 translocations and/or transposable element (TE) movement. Translocations have been 301 shown to increase Y-chromosome content2,55. In contrast, although TE movement has 302 historically been considered to be a deleterious process, more recent reports have 303 revealed that TEs can alter regions by removing or adding regulatory or coding 304 sequence39,56-59, and TEs may even act as substrate for novel genes60. As predicted for 305 non-recombining regions, we found a large number of TEs in the five P. parae Y 306 chromosomes.

307

308 Making and Maintaining Five Morphs

309 We found remarkable morph specific genetic diversity on the Y chromosome of P. 310 parae. Intriguingly, that diversity is maintained within populations, as evidenced by the 311 stability in morph frequencies over repeated surveys spanning 50 years, or roughly 150 312 generations13,14,20. Even if alternative morphs have exactly equal fitness, populations 313 are expected to eventually fix for one morph due to drift61. Maintaining alternative 314 morphs within the same population relies on negative frequency dependent selection, 315 thus as one morph decreases in frequency its fitness increases, such as with the three 316 male morphs of the side blotched lizard62.

317 Previous work suggests that P. parae morphs are also under negative frequency 318 dependent selection14-17, and this could facilitate the establishment and maintenance of 319 five distinct Y chromosomes within the same species. Most new mutations are expected 320 to be lost through drift if they do not confer a high enough fitness advantage over 321 alternative alleles63, but mutations resulting in a new morph would be at the lowest 322 frequencies and thus have the highest fitness, allowing them to rapidly stabilize in the 323 population. Alternatively, it is possible that the morphs arose in separate populations 324 and only later came into sympatry.

325

326 Genetic Basis of Male Reproductive Morphs

327 Autosomal non-recombining regions have been shown to be associated with alternative 328 reproductive strategies in a range of species7,9,64-68. For example, male morphs of the 329 white throated sparrow, which differ in pigmentation and social behavior, are the result 330 of a hybridization event which instantly brought together alternative sequence and 331 halted recombination65. Importantly, because none of the characteristics of the P. parae 332 morphs are found in any close relatives, it is unlikely that hybridization is the source of 333 the Y chromosomes we describe.

334 The alternative male morphs in the ruff are controlled by an autosomal supergene that 335 is composed of two alternative versions of an inverted region10. It has yet to be 336 determined whether the diversity across these ruff supergenes pre-dates the inversion 337 or arose after recombination stopped as it did in P. parae.

338 A large inverted region is also associated with social morphs in many ant species, this 339 region formed in a common ancestor and has been maintained by balancing selection 340 through repeated speciation events12. Although it is possible that the multiple male 341 morphs in P. parae arose in an ancestor, this is less likely as they have not been 342 observed in any related species to date.

343

344 Conclusion

345 The role of recombination in shaping co-adapted allele complexes has long been an 346 enigma, given that recombination is both a key mechanism in generating diverse allelic 347 combinations, yet recombination also acts to break up such combinations. Here we 348 found that tremendous diversity can still be generated without the power of 349 recombination, and the Y chromosome contains remarkable adaptive potential with 350 regard to male phenotypic evolution. The five Y-linked male morphs of P. parae 351 emerged and diverged after recombination was halted, resulting in five unique Y 352 chromosomes within one species. Future work identifying the mechanisms by which 353 morphs are determined by these five Y chromosomes will provide much needed insight

354 to determining which evolutionary forces have led to and shaped these amazing 355 complexes and their co-evolution with the rest of the genome, which is shared across all 356 morphs.

357

358 Methods

359

360 Field Collections and DNA Isolation

361 To ensure we accounted for natural diversity in the five Y chromosomes of P. parae, 362 and because this species is extremely difficult to breed in captivity, we collected all 363 samples (N=40) from large native populations around Georgetown, Guyana in 2016 364 (see Table S2 and Sandkam, et al. 18 for description of populations) (Environmental 365 Protection Agency of Guyana Permit 120616 SP: 015). Individuals were rapidly 366 sacrificed in MS-222, whole-tail tissue was dissected into EtOH and immediately placed 367 in liquid nitrogen to maintain integrity of high molecular weight (HMW) DNA. Tissue 368 samples were brought back to the lab and kept at -80º C until HMW DNA extraction.

369 HMW DNA was extracted from 25mg tail tissue of each sample following a modified 370 protocol from 10X Genomics described in Almeida et al24. Briefly, nuclei were isolated 371 by gently homogenizing tissue with a pestle in cold Nuclei Isolation Buffer from a Nuclei 372 PURE Prep Kit (Sigma). Nuclei were pelleted and supernatant removed before being 373 digested by incubating in 70 µl PBS, 10 µl Proteinase K (Qiagen), and 70 µl Digestion 374 buffer (20mM EDTA, 2mM Tris-HCL, 10mM N-Laurylsarcosine) for two hours at room 375 temperature on a tube rotator. Tween 20 was added (0.1% final concentration) and 376 DNA was bound to SPRIselect magnetic beads (Beckman Coulter) for 20 min. Beads 377 were bound to a magnetic rack and washed twice with 70% Ethanol before eluting DNA. 378 Samples were visually screened for integrity of HMW DNA on an agarose gel. Of the 40 379 samples extracted, 29 passed initial screening and were used for individual 10X 380 Chromium linked-read sequencing (six female, seven red, five yellow, one blue, four 381 immaculata, and six parae morph). The remaining 11 samples (five female, one red, 382 three blue, one immaculata, and one parae morph) were individually sequenced on an

383 Illumina HiSeqX as 2 x 150 bp reads using the v2.5 sequencing chemistry with 300bp 384 inserts and trimmed with trimmomatic (v0.36)69. To ensure high coverage of the Y, all 40 385 samples were individually sequenced to a predicted coverage of 40X (see Table S2 for 386 number of reads after filtering), which would result in predicted 20X coverage of the 387 haploid Y.

388

389 10X Chromium Linked-read Sequencing and Assembly

390 10X Chromium linked-read sequencing was performed at the SciLifeLab, Uppsala 391 Sweden. The 10X Chromium pipeline adds unique tags to each piece of HMW DNA 392 before sequencing on an Illumina platform. These tagged reads were either used 393 directly in the 10X assembly pipeline, or tags were removed using the basic function of 394 Longranger v.2.2.2 (10X Genomics), trimmed with trimmomatic (v0.36)69 and treated as 395 normal Illumina reads70 for coverage analyses (Table S2).

396 Scaffold level de novo genomes were assembled for each of the 29 linked-read 397 samples using the Supernova v2.1.1 software package (10X Genomics) (see Table S3 398 for assembly statistics).

399 A female chromosome level assembly was created by assigning scaffolds to 400 chromosomal positions. The female with the best de novo assembly (largest assembly 401 size and scaffold N50) was used for the Reference Assisted Chromosome Assembly 402 (RACA) pipeline31. Briefly, scaffolds were aligned with LASTZ v1.0471 against high- 403 quality chromosome level genome assemblies of a close relative (Xiphophorus helleri 404 v4.0; GenBank accession GCA_003331165) and an outgroup (Oryzias latipes v1; 405 GenBank accession GCA_002234675). Alignments were then run through the UCSC 406 chains and nets pipeline from the kentUtils software suite72 before passing to the RACA 407 pipeline. RACA uses alignments of short-insert and long-insert paired reads that bridge 408 scaffolds to further order and confirm scaffold arrangement. For short-insert data, 150bp 409 reads from the five females sequenced with paired-end Illumina (300bp inserts) were 410 aligned to the target assembly with Bowtie2 v2.2.973 reporting concordant mappings 411 only (--no-discordant option). For long-insert data, synthetic 150bp ‘pseudo-mate-pair’ 412 reads were generated from the de novo scaffolds of the 6 female P. parae samples

413 sequenced with Chromium linked-reads. To increase the likelihood that bridge reads 414 spanned scaffolds, we generated two long-insert pseudo-mate-pair libraries for each of 415 the six de novo female genomes, a 2.3kb insert library and a 15kb insert library, and 416 aligned these to the target assembly. RACA then used the information from both the 417 phylogenetically weighted genome pairwise alignments, and the read mapping data to 418 order the target scaffolds into longer predicted chromosome fragments.

419 To identify which chromosome is the sex chromosome and determine the extent of X-Y 420 divergence, we mapped reads from all 40 individuals to the female scaffolds that had 421 RACA generated chromosome annotations using the aln function of bwa (v0.6.1)74. 422 Alignments were filtered for uniquely mapped reads and average scaffold coverage was 423 calculated using soap.coverage v2.7.7 (http://soap.genomics.org.cn/). To account for 424 differences across individuals in sequencing library size, we divided the coverage of 425 each scaffold by the average coverage across all scaffolds for each individual. Male to 426 female (M:F) fold change in coverage of each scaffold was calculated for all males, and

427 each of the five morphs as log2(average male coverage) – log2(average female 428 coverage)75. Upon observing chromosome 8 was the sex chromosome, we calculated 429 the 95% CI for M:F coverage by bootstrapping across all scaffolds which RACA placed 430 on the autosomes.

431

432 Identifying Morph specific sequence by k-mers

433 To locate morph specific sequence, we identified morph specific k-mers (morph-mers) 434 and then mapped these to the respective de novo genome assemblies. First Jellyfish 435 v2.2.374 was used to identify all 31bp k-mers from the ‘megabubble’ output of each of 436 the 22 male de novo genome assemblies from supernova. The ‘megabubble’ outputs 437 from supernova include all sequence before being flattened into final phased scaffolds, 438 this minimizes any potential k-mer loss due to arbitrary flattening (a benefit over k-mers 439 from the full psuedohap output of supernova) and excludes k-mers from sequencing 440 errors (a benefit over k-mers from sequencing reads). We next identified the putative Y- 441 linked k-mers in each sample by removing all k-mers present in any of the females. For 442 female k-mer identification we conservatively took k-mers from both the megabubble

443 outputs and the raw Illumina reads of all 11 female individuals, which we used to identify 444 every 31bp k-mer present >3 times (this maximized the chance of including k-mers from 445 unassembled regions of female genomes but minimized k-mers from sequencing 446 errors35). The k-mers present in females all occur either on autosomes or the X 447 chromosome, therefore by removing the female k-mers from all k-mers identified in 448 males we are left with putative Y-linked k-mers that we call Y-mers21,35. To exclude k- 449 mers representing autosomal SNPs unique to a single male we required Y-mers be 450 present in at least two individuals. Since all female sequence is theoretically present in 451 males, we validated our method by identifying all k-mers from the six female 452 megabubbles that were not present in the male Illumina reads and found how many 453 were present in at least two female individuals.

454 We then combined all the Y-mers we found with those we previously identified in P. 455 picta21 and used the presence of Y-mers as character states to build a phylogeny of all 456 individuals and P. picta. Two runs of MrBayes v3.2.276 were run for 100,000 generations 457 with Y-mers treated akin to restriction sites (model F81 with rates set to equal). The SD 458 of the split frequencies between runs reached 0 indicating both runs converged on 459 identical and robust trees.

460 Monophyletic clades were recovered for each of the major morphs (immaculata, parae 461 and melanzona). We then identified unique Y-mers in each clade (Y-mers present in 462 every individual of that clade but not present outside that clade). This approach 463 provided us with all morph-mers (Y-mers present in every individual of a given morph 464 but not present in any individual of the other morphs).

465 These morph-mers reveal two insights: (1) at a gross level they provide a sense of how 466 much Y sequence is shared within versus across morphs and (2) mapping these morph- 467 mers to the respective de novo genomes allows us to identify regions of morph specific 468 sequence34,35,77,78. To find these regions of morph specific sequence we first mapped 469 the corresponding set of 31bp morph-mers to each of the 22 de novo male genomes 470 (pseudohap style of Supernova output) using bowtie273 allowing for no mismatches, 471 gaps, or trimming. We found morph-mers disproportionately map to scaffolds, indicating

472 they came from regions of highly diverged morph-specific sequence rather than evenly 473 dispersed lowly diverged sequence (Fig. S3 and S4).

474 To verify our pipeline was identifying true Y-specific alignments we attempted to align 475 the Y-mers and all of the morph-mers to each of the six female de novo genome 476 assemblies and found no alignments could be made. We next verified that our morph- 477 mers were targeting morph specific sequence by attempting to align each of the morph- 478 mer datasets to individuals of the opposite morphs and again found no alignments could 479 be made.

480

481 Coverage analysis

482 To independently verify that the scaffolds identified by our k-mer approach contained 483 highly diverged sequence, we aligned each of the 39 individuals to the individual with 484 the best de novo genome of each morph (based on assembly size, scaffold N50, and 485 contig N50, Supplementary Table 3). Alignments were generated with the mem function 486 of bwa (v0.7.17)74. samtools (v1.10) was used to remove unmapped reads and 487 secondary alignments with the fixmate function, and duplicates were removed with 488 markdup. bamqc was then used to assess distribution of map quality. For each 489 individual, the average coverage of each scaffold by reads with mapq >60 was 490 determined using the depth function of samtools (v1.10). Y chromosomes are notorious 491 for high incidents of transposable elements and repeats, this highly conservative filtering 492 decreased false alignments to these regions. To account for differences across 493 individuals in sequencing library size, we took the coverage of each scaffold divided by 494 that individual’s coverage across all scaffolds. The average raw scaffold coverage 495 across all individuals was 29.97X, therefore any scaffold with a corrected coverage < 496 0.025 (raw coverage <1X) was considered to have a coverage of 0.

497

498 Gene Annotation

499 To identify genes on morph specific scaffolds we followed the pipeline described in 500 Almeida et al24. Briefly, we took a very conservative approach by annotating only the

501 scaffolds with >5 morph-mers from each of the de novo references used for the 502 coverage analysis (one of each morph). The chance of a scaffold containing any 503 particular 31bp k-mer depends on the length of the scaffold and can be calculated 504 roughly as 0.2531 x scaffold length. The longest male scaffold we recovered was 505 19,887,348 and therefore had the greatest chance of containing any given Y-mer; 506 4.31*10-12. The most abundant morph-mers were melanzona-mers (87,629), therefore 507 the likelihood of the largest scaffold containing one melanzona-mer by chance was 508 4.31*10-12 x 87,629 = 3.78*10-7 and the likelihood it contains 5 melanzona-mers purely 509 by chance was roughly 7.71*10-33. If we conservatively assume that all scaffolds have 510 the same probability of containing a Y-mer by chance and the male with the most 511 scaffolds had 25,416 scaffolds – there was a likelihood of 1.96*10-28 that a scaffold was 512 incorrectly identified.

513 We then annotated these scaffolds with MAKER v2.31.1079. We ran the MAKER 514 pipeline twice: first based on a guppy-specific repeat library, protein sequence, EST and 515 RNA sequence data (later used to train ab-initio software) and a second time combining 516 evidence data from the first run and ab-initio predictions. We create a repeat library for 517 these scaffolds using de novo repeats identified by RepeatModeler v1.0.1080 which we 518 then combined with Actinopterygii-specific repeats to use with RepeatMasker v4.0.781. 519 Annotated protein sequences were downloaded from Ensembl (release 95)82 for 8 fish 520 species: Danio rerio (GRCz11), Gasterosteus aculeatus (BROADS1), Oryzias latipes 521 (ASM223467v1), Poecilia latipinna (1.0), Poecilia mexicana (1.0), Poecilia reticulata 522 (1.0), Takifugu rubripes (FUGU5) and Xiphophorus maculatus (5.0). For ESTs, we used 523 10,664 tags isolated from guppy embryos and male testis83. Furthermore, to support 524 gene predictions we also used two publicly available libraries of RNA-seq data collected 525 from guppy male testis and male embryos84 and assembled with StringTie 1.3.3b85. As 526 basis for the construction of gene models, we combined ab-initio predictions from 527 Augustus v3.2.386, trained via BUSCO v3.0.287, and SNAP v2006-07-2888. To train 528 Augustus and SNAP, we ran the MAKER pipeline a first time to create a profile using 529 the protein and EST evidence along with RNA-seq data. Both Augustus and SNAP were 530 then trained from this initial evidence-based annotation. Functional inference for genes 531 and transcripts was performed using the translated CDS features of each coding

532 transcript. Gene names and protein functions were retrieved using BLASTp to search 533 the Uniprot/Swissprot, InterProscan v5 and GenBank databases.

534

535 Data Availability

536 All reads will be submitted to GenBank.

537

538 Acknowledgements

539 We thank the members of the Mank lab and Dr. Nora Prior for stimulating conversations 540 and excellent feedback on early drafts of the manuscript. This was supported by the 541 Natural Sciences and Engineering Research Council of Canada through a Banting 542 Postdoctoral Fellowship (to B.A.S.), the European Research Council (Grants 260233 543 and 680951 to J.E.M.), and a Canada 150 Research Chair (to J.E.M.). Field work was 544 conducted under Permit 120616 SP: 015 from the Environmental Protection Agency of 545 Guyana. Sequencing was performed by the SNP&SEQ Technology Platform in 546 Uppsala, Sweden. The CEIBA Biological Center partially subsidized our expenses 547 during field collection in Guyana. We thank Clara Lacy for the illustrations of P. parae.

548

549 Figure 1. Coverage differences between the sexes (male:female log2) for female 550 scaffolds placed by RACA on the reference Xiphophorus hellerii chromosomes. (A) 551 Average immaculata (inner ring), the three melanzona (middle ring) and parae morphs 552 (outer ring) plotted across all chromosomes. Highlighted in blue is X. hellerii 553 chromosome 8 which is syntenic to the guppy sex chromosome (P. reticulata 554 chromosome 12). The decreased male coverage of chromosome 8 indicates this is also 555 the sex chromosome in P. parae. (B) All five P. parae morphs share the same pattern of 556 XY divergence, indicating a shared history of recombination suppression. (C) Pattern of 557 P. parae XY divergence is the same as the sister species P. picta, indicating 558 recombination was stopped in the common ancestor of P. parae and P. picta (14.8-18.5 559 mya43). In each, horizontal grey-shaded areas represent the 95% confidence intervals

560 based on bootstrap estimates across the autosomes.

561 562 Figure 2. Bayesian Y chromosome phylogeny based on presence/absence of the 563 27,950,090 P. parae Y-mers and 1,646 P. picta Y-mers21 in each individual and rooted 564 on P. picta. The posterior probability is presented above each node, below the node is 565 the number of P. parae Y-mers unique to all members of that clade. The three major 566 morphs of P. parae (immaculata, parae and melanzona) formed distinct clades and the 567 Y-mers unique to all members of these clades are called morph-mers.

568 569 Figure 3. The distribution of the 27,950,090 P. parae Y-mers reveals strong differences 570 across morphs. While there are very few Y-mers present in all morphs, each morph 571 harbors unique Y-mers. The melanzona and parae morphs share more Y-mers with one 572 another than either share with immaculata.

573 574 Figure 4. Relative corrected scaffold coverage of 39 individuals when aligned to 575 melanzona (A-B), immaculata (C-D), and parae (E-F) de novo genomes. Scaffolds 576 containing morph-mers had higher coverage by males than females (A,C,E) confirming 577 these scaffolds contain male specific sequence. Scaffolds containing morph-mers also

578 had higher coverage by males of the reference morph than males of the other morphs 579 (B,D,F), indicating the Y chromosome sequence is substantially diverged across 580 morphs. In each, corrected scaffold coverage of focal morph is on the Y axis and 581 corrected scaffold coverage of the compared morph is on the X axis. The 1:1 line is 582 denoted as a grey dashed line. The linear regression and standard error across all 583 scaffolds are shown as a thick black line that is nearly 1:1 for all morphs (note – 95% 584 confidence interval is presented but too small to distinguish from the regression line). 585 The scaffolds containing morph-mers are shown as colored points, the linear regression 586 and standard error of morph-mer scaffolds are shown as a colored line and shaded 587 region respectively.

588 References

589 590 1 Bachtrog, D. Y-chromosome evolution: emerging insights into processes of Y- 591 chromosome degeneration. Nat. Rev. Genet. 14, 113-124, doi:10.1038/nrg3366 592 (2013). 593 2 Tobler, R., Nolte, V. & Schlotterer, C. High rate of translocation-based gene birth 594 on the Drosophila Y chromosome. P Natl Acad Sci USA 114, 11721-11726, 595 doi:10.1073/pnas.1706502114 (2017). 596 3 Mahajan, S. & Bachtrog, D. Convergent evolution of Y chromosome gene 597 content in flies. Nat Commun 8, 785, doi:10.1038/s41467-017-00653-x (2017). 598 4 Bachtrog, D., Mahajan, S. & Bracewell, R. Massive gene amplification on a 599 recently formed Drosophila Y chromosome. Nat Ecol Evol 3, 1587-1597, 600 doi:10.1038/s41559-019-1009-9 (2019). 601 5 Hall, A. B. et al. Radical remodeling of the Y chromosome in a recent radiation of 602 malaria mosquitoes. P Natl Acad Sci USA 113, E2114-2123, 603 doi:10.1073/pnas.1525164113 (2016). 604 6 Todesco, M. et al. Massive haplotypes underlie ecotypic differentiation in 605 sunflowers. Nature, doi:10.1038/s41586-020-2467-6 (2020). 606 7 Schwander, T., Libbrecht, R. & Keller, L. Supergenes and Complex Phenotypes. 607 Curr. Biol. 24, R288-R294, doi:10.1016/j.cub.2014.01.056 (2014). 608 8 Wang, J. et al. A Y-like social chromosome causes alternative colony 609 organization in fire ants. Nature 493, 664-668, doi:10.1038/nature11832 (2013). 610 9 Lamichhaney, S. et al. Structural genomic changes underlie alternative 611 reproductive strategies in the ruff (Philomachus pugnax). Nat. Genet. 48, 84-88, 612 doi:10.1038/ng.3430 (2016). 613 10 Kupper, C. et al. A supergene determines highly divergent male reproductive 614 morphs in the ruff. Nat. Genet. 48, 79-83, doi:10.1038/ng.3443 (2016). 615 11 Branco, S. et al. Multiple convergent supergene evolution events in mating-type 616 chromosomes. Nature Communications 9, 2000, doi:10.1038/s41467-018-04380- 617 9 (2018). 618 12 Yan, Z. et al. Evolution of a supergene that regulates a trans-species social 619 polymorphism. Nat Ecol Evol 4, 240-249, doi:10.1038/s41559-019-1081-1 620 (2020). 621 13 Lindholm, A. K., Brooks, R. & Breden, F. Extreme polymorphism in a Y-linked 622 sexually selected trait. Heredity (Edinb) 92, 156-162, doi:10.1038/sj.hdy.6800386 623 (2004). 624 14 Hurtado-Gonzales, J. L. & Uy, J. A. Alternative mating strategies may favour the 625 persistence of a genetically based colour polymorphism in a pentamorphic fish. 626 Anim. Behav. 77, 1187-1194, doi:10.1016/j.anbehav.20 (2009). 627 15 Hurtado-Gonzales, J. L. & Uy, J. A. Intrasexual competition facilitates the 628 evolution of alternative mating strategies in a colour polymorphic fish. BMC Evol. 629 Biol. 10, 391, doi:10.1186/1471-2148-10-391 (2010). 630 16 Hurtado-Gonzales, J. L., Baldassarre, D. T. & Uy, J. A. Interaction between 631 female mating preferences and predation may explain the maintenance of rare

632 males in the pentamorphic fish Poecilia parae. J. Evol. Biol. 23, 1293-1301, 633 doi:10.1111/j.1420-9101.2010.01995.x (2010). 634 17 Hurtado-Gonzales, J. L., Loew, E. R. & Uy, J. A. Variation in the visual habitat 635 may mediate the maintenance of color polymorphism in a poeciliid fish. PLoS 636 One 9, e101497, doi:10.1371/journal.pone.0101497 (2014). 637 18 Sandkam, B. A., Young, C. M., Breden, F. M., Bourne, G. R. & Breden, F. Color 638 vision varies more among populations than among species of live-bearing fish 639 from South America. BMC Evol. Biol. 15, 225, doi:10.1186/s12862-015-0501-3 640 (2015). 641 19 Bourne, G. R., Breden, F. & Allen, T. C. Females prefer carotenoid colored males 642 as mates in the pentamorphic livebearing fish, Poecilia parae. 643 Naturwissenschaften 90, 402-405, doi:10.1007/s00114-003-0444-1 (2003). 644 20 Liley, N. R. Reproductive Isolation in some sympatric species of fishes Doctor of 645 Philosophy thesis, Oxford, (1963). 646 21 Darolti, I. et al. Extreme heterogeneity in sex chromosome differentiation and 647 dosage compensation in livebearers. Proceedings of the National Academy of 648 Sciences 116, 19031-19036, doi:10.1073/pnas.1905298116 (2019). 649 22 Wright, A. et al. Convergent recombination suppression suggests a role of sexual 650 selection in guppy sex chromosome formation. Nature Communications 8, 14251 651 (2017). 652 23 Darolti, I., Wright, A. E. & Mank, J. E. Guppy Y Chromosome Integrity Maintained 653 by Incomplete Recombination Suppression. Genome Biol Evol 12, 965-977, 654 doi:10.1093/gbe/evaa099 (2020). 655 24 Almeida, P. et al. Divergence and Remarkable Diversity of the Y Chromosome in 656 Guppies. bioRxiv, doi:10.1101/2020.07.13.200196 (2020). 657 25 Reznick, D. N., Miles, D. B. & Winslow, S. Life History of Poecilia picta 658 (Poeciliidae) from the Island of Trinidad. Copeia 1992, 782-790, 659 doi:10.2307/1446155 (1992). 660 26 Haskins, C. P. & Haskins, E. F. The role of sexual selection as an isolating 661 mechanism in three species of Poeciliid fishes. Evolution 3, 160-169 (1949). 662 27 Liley, N. R. Ethological Isolating Mechanisms in Four Sympatric Species of 663 Poeciliid Fishes. Behaviour Supplement 13, 1-197 (1966). 664 28 Vicoso, B. & Bachtrog, D. Reversal of an ancient sex chromosome to an 665 autosome in Drosophila. Nature 499, 332-335, doi:10.1038/nature12235 (2013). 666 29 Vicoso, B. & Bachtrog, D. Numerous transitions of sex chromosomes in Diptera. 667 PLoS Biol. 13, e1002078, doi:10.1371/journal.pbio.1002078 (2015). 668 30 Vicoso, B., Emerson, J. J., Zektser, Y., Mahajan, S. & Bachtrog, D. Comparative 669 sex chromosome genomics in snakes: differentiation, evolutionary strata, and 670 lack of global dosage compensation. PLoS Biol. 11, e1001643, 671 doi:10.1371/journal.pbio.1001643 (2013). 672 31 Kim, J. et al. Reference-assisted chromosome assembly. 110, 1785-1790, 673 doi:10.1073/pnas.1220349110 %J Proceedings of the National Academy of 674 Sciences (2013). 675 32 Pucholt, P., Wright, A. E., Conze, L. L., Mank, J. E. & Berlin, S. Recent Sex 676 Chromosome Divergence despite Ancient Dioecy in the Willow Salix viminalis. 677 Mol. Biol. Evol. 34, 1991-2001, doi:10.1093/molbev/msx144 (2017).

678 33 Akagi, T., Henry, I. M., Tao, R. & Comai, L. A Y-chromosome-encoded small 679 RNA acts as a sex determinant in persimmons. Science 346 (2014). 680 34 Torres, M. F. et al. Genus-wide sequencing supports a two-locus model for sex- 681 determination in Phoenix. Nat Commun 9, 3969, doi:10.1038/s41467-018-06375- 682 y (2018). 683 35 Morris, J., Darolti, I., Bloch, N. I., Wright, A. E. & Mank, J. E. Shared and 684 Species-Specific Patterns of Nascent Y Chromosome Evolution in Two Guppy 685 Species. Genes (Basel) 9, doi:10.3390/genes9050238 (2018). 686 36 Napolitano, L. M. & Meroni, G. TRIM family: Pleiotropy and diversification 687 through homomultimer and heteromultimer formation. IUBMB Life 64, 64-71, 688 doi:10.1002/iub.580 (2012). 689 37 Sardiello, M., Cairo, S., Fontanella, B., Ballabio, A. & Meroni, G. Genomic 690 analysis of the TRIM family reveals two groups of genes with distinct evolutionary 691 properties. BMC Evol. Biol. 8, 225, doi:10.1186/1471-2148-8-225 (2008). 692 38 Karki, R. et al. NLRC3 is an inhibitory sensor of PI3K-mTOR pathways in cancer. 693 Nature 540, 583-587, doi:10.1038/nature20597 (2016). 694 39 Sandkam, B. A. et al. Tbx2a Modulates Switching of RH2 and LWS Opsin Gene 695 Expression. Mol. Biol. Evol. 37, 2002-2014, doi:10.1093/molbev/msaa062 (2020). 696 40 Showell, C., Christine, K. S., Mandel, E. M. & Conlon, F. L. Developmental 697 expression patterns of Tbx1, Tbx2, Tbx5, and Tbx20 in Xenopus tropicalis. Dev. 698 Dyn. 235, 1623-1630, doi:10.1002/dvdy.20714 (2006). 699 41 Gibson-Brown, J. J., S, I. A., Silver, L. M. & Papaioannou, V. E. Expression of T- 700 box genes Tbx2-Tbx5 during chick organogenesis. Mech Dev 74, 165-169 701 (1998). 702 42 Tomaszkiewicz, M., Chalopin, D., Schartl, M., Galiana, D. & Volff, J.-N. A 703 multicopy Y-chromosomal SGNH hydrolase gene expressed in the testis of the 704 platyfish has been captured and mobilized by a Helitron transposon. BMC Genet. 705 15 (2014). 706 43 Rabosky, D. L. et al. An inverse latitudinal gradient in speciation rate for marine 707 fishes. Nature 559, 392-395, doi:10.1038/s41586-018-0273-1 (2018). 708 44 Felsenstein, J. The evolutionary advantage of recombination. Genetics 78, 737- 709 756 (1974). 710 45 Lenormand, T., Fyon, F., Sun, E. & Roze, D. Sex Chromosome Degeneration by 711 Regulatory Evolution. Curr. Biol., doi:10.1016/j.cub.2020.05.052 (2020). 712 46 Wright, A. E., Dean, R., Zimmer, F. & Mank, J. E. How to make a sex 713 chromosome. Nat Commun 7, 12087, doi:10.1038/ncomms12087 (2016). 714 47 Hough, J., Wang, W., Barrett, S. C. H. & Wright, S. I. Hill-Robertson Interference 715 Reduces Genetic Diversity on a Young Plant Y-Chromosome. Genetics 207, 685 716 (2017). 717 48 Marais, G. A. et al. Evidence for degeneration of the Y chromosome in the 718 dioecious plant Silene latifolia. Curr. Biol. 18, 545-549, 719 doi:10.1016/j.cub.2008.03.023 (2008). 720 49 Bachtrog, D. & Charlesworth, B. Reduced adaptation of a non-recombining neo- 721 Y chromosome. Nature 416, 323-326, doi:10.1038/416323a (2002).

722 50 Mank, J. E. Small but mighty: the evolutionary dynamics of W and Y sex 723 chromosomes. Chromosome Res 20, 21-33, doi:10.1007/s10577-011-9251-2 724 (2012). 725 51 Kaiser, V. B. & Charlesworth, B. Muller's ratchet and the degeneration of the 726 Drosophila miranda neo-Y chromosome. Genetics 185, 339-348, 727 doi:10.1534/genetics.109.112789 (2010). 728 52 Keightley, P. D. & Otto, S. P. Interference among deleterious mutations favours 729 sex and recombination in finite populations. Nature 443, 89-92, 730 doi:10.1038/nature05049 (2006). 731 53 Ritz, K. R., Noor, M. A. F. & Singh, N. D. Variation in Recombination Rate: 732 Adaptive or Not? Trends Genet. 33, 364-374, doi:10.1016/j.tig.2017.03.003 733 (2017). 734 54 Ellegren, H. Characteristics, causes and evolutionary consequences of male- 735 biased mutation. Proc Biol Sci 274, 1-10, doi:10.1098/rspb.2006.3720 (2007). 736 55 Tennessen, J. A. et al. Repeated translocation of a gene cassette drives sex- 737 chromosome turnover in strawberries. PLoS Biol. 16, e2006062, 738 doi:10.1371/journal.pbio.2006062 (2018). 739 56 Bourque, G. et al. Ten things you should know about transposable elements. 740 Genome Biol 19, 199, doi:10.1186/s13059-018-1577-z (2018). 741 57 Carleton, K. L. et al. Movement of transposable elements contributes to cichlid 742 diversity. doi:10.1101/2020.02.26.961987 (2020). 743 58 Brawand, D. et al. The genomic substrate for adaptive radiation in African cichlid 744 fish. Nature 513, 375-+, doi:10.1038/nature13726 (2014). 745 59 Auvinet, J. et al. Mobilization of retrotransposons as a cause of chromosomal 746 diversification and rapid speciation: the case for the Antarctic teleost genus 747 Trematomus. BMC Genomics 19, 339, doi:10.1186/s12864-018-4714-x (2018). 748 60 Naville, M. et al. Not so bad after all: retroviruses and long terminal repeat 749 retrotransposons as a source of new genes in vertebrates. Clin. Microbiol. Infect. 750 22, 312-323, doi:10.1016/j.cmi.2016.02.001 (2016). 751 61 Sinervo, B. & Calsbeek, R. The Developmental, Physiological, Neural, and 752 Genetical Causes and Consequences of Frequency-Dependent Selection in the 753 Wild. Annual Review of Ecology, Evolution, and Systematics 37, 581-610, 754 doi:10.1146/annurev.ecolsys.37.091305.110128 (2006). 755 62 Sinervo, B. & Lively, C. M. The rock-paper-scissors game and the evolution of 756 alternative male strategies. Nature 380, 240-243, doi:DOI 10.1038/380240a0 757 (1996). 758 63 Hartl, D. L. & Clark, A. G. Principles of Population Genetics. (Sinauer 759 Associates, 1997). 760 64 Lank, D. B., Smith, C. M., Hanotte, O., Burke, T. & Cooke, F. Genetic 761 polymorphism for alternative mating behaviour in lekking male ruff Philomachus 762 pugnax. Nature 378, 59-62, doi:DOI 10.1038/378059a0 (1995). 763 65 Tuttle, E. M. et al. Divergence and Functional Degradation of a Sex 764 Chromosome-like Supergene. Curr. Biol. 26, 344-350, 765 doi:10.1016/j.cub.2015.11.069 (2016). 766 66 Hunt, B. G. Supergene Evolution: Recombination Finds a Way. Curr. Biol. 30, 767 R73-R76, doi:10.1016/j.cub.2019.12.006 (2020).

768 67 Charlesworth, D. The status of supergenes in the 21st century: recombination 769 suppression in Batesian mimicry and sex chromosomes and other complex 770 adaptations. Evol Appl 9, 74-90, doi:10.1111/eva.12291 (2016). 771 68 Joron, M. et al. Chromosomal rearrangements maintain a polymorphic 772 supergene controlling butterfly mimicry. Nature 477, 203-206, 773 doi:10.1038/nature10341 (2011). 774 69 Bolger, A. M., Lohse, M. & Usadel, B. Trimmomatic: a flexible trimmer for Illumina 775 sequence data. Bioinformatics 30, 2114-2120, doi:10.1093/bioinformatics/btu170 776 (2014). 777 70 Elyanow, R., Wu, H.-T. & Raphael, B. J. Identifying structural variants using 778 linked-read sequencing data. Bioinformatics (Oxford, England) 34, 353-360, 779 doi:10.1093/bioinformatics/btx712 (2018). 780 71 Harris, R. S. Improved pairwise alignment of genomic DNA Doctor of Philosophy 781 thesis, Penn State, (2007). 782 72 Kent, W. J., Baertsch, R., Hinrichs, A., Miller, W. & Haussler, D. Evolution's 783 cauldron: duplication, deletion, and rearrangement in the mouse and human 784 genomes. P Natl Acad Sci USA 100, 11484-11489, 785 doi:10.1073/pnas.1932072100 (2003). 786 73 Langmead, B. & Salzberg, S. L. Fast gapped-read alignment with Bowtie 2. Nat. 787 Methods 9, 357 (2012). 788 74 Li, H. & Durbin, R. Fast and accurate long-read alignment with Burrows–Wheeler 789 transform. Bioinformatics 26, 589-595, doi:10.1093/bioinformatics/btp698 (2010). 790 75 Palmer, D. H., Rogers, T. F., Dean, R. & Wright, A. E. How to identify sex 791 chromosomes and their turnover. Mol. Ecol. 28, 4709-4724, 792 doi:10.1111/mec.15245 (2019). 793 76 Ronquist, F. et al. MrBayes 3.2: efficient Bayesian phylogenetic inference and 794 model choice across a large model space. Syst. Biol. 61, 539-542, 795 doi:10.1093/sysbio/sys029 (2012). 796 77 Carvalho, A. B. & Clark, A. G. Efficient identification of Y chromosome 797 sequences in the human and Drosophila genomes. Genome Res. 23, 1894- 798 1907, doi:10.1101/gr.156034.113 (2013). 799 78 Carvalho, A. B., Vicoso, B., Russo, C. A., Swenor, B. & Clark, A. G. Birth of a 800 new gene on the Y chromosome of Drosophila melanogaster. P Natl Acad Sci 801 USA 112, 12450-12455, doi:10.1073/pnas.1516543112 (2015). 802 79 Holt, C. & Yandell, M. MAKER2: an annotation pipeline and genome-database 803 management tool for second-generation genome projects. BMC Bioinformatics 804 12, 491, doi:10.1186/1471-2105-12-491 (2011). 805 80 RepeatModeler Open-1.0 (http://www.repeatmasker.org, 2015). 806 81 RepeatMasker Open-4.0 (http://www.repeatmasker.org, 2015). 807 82 Howe, K. L. et al. Ensembl Genomes 2020-enabling non-vertebrate genomic 808 research. Nucleic Acids Res. 48, D689-D695, doi:10.1093/nar/gkz890 (2020). 809 83 Dreyer, C. et al. ESTs and EST-linked polymorphisms for genetic mapping and 810 phylogenetic reconstruction in the guppy, Poecilia reticulata. BMC Genomics 8, 811 269, doi:10.1186/1471-2164-8-269 (2007).

812 84 Sharma, E. et al. Transcriptome assemblies for studying sex-biased gene 813 expression in the guppy, Poecilia reticulata. BMC Genomics 15, 400, 814 doi:10.1186/1471-2164-15-400 (2014). 815 85 Pertea, M. et al. StringTie enables improved reconstruction of a transcriptome 816 from RNA-seq reads. Nat. Biotechnol. 33, 290-295, doi:10.1038/nbt.3122 (2015). 817 86 Stanke, M. et al. AUGUSTUS: ab initio prediction of alternative transcripts. 818 Nucleic Acids Res. 34, W435-439, doi:10.1093/nar/gkl200 (2006). 819 87 Seppey, M., Manni, M. & Zdobnov, E. M. in Gene Prediction 227-245 820 (Springer, 2019). 821 88 Korf, I. Gene finding in novel genomes. BMC Bioinformatics 5, 59, 822 doi:10.1186/1471-2105-5-59 (2004). 823